Ricardian Models Kim J.

Ruhl

Ricardian Models
[My notes are in beta. If you see something that doesn’t look right, I would greatly appreciate a heads-up.]
In Ricardian models, first laid out in Ricardo (1817), each country in the economy can produce
each good, but countries differ in how well they can produce each good. Trade arises to exploit
comparative advantage: A country exports the goods it is relatively good at producing and im-
ports goods that other countries are relatively good at producing.

1 Two countries and a continuum of goods

The simplest Ricardian model is the two-good two-country example from undergrad textbooks,
but moving to a continuum of goods isn’t much more difficult, and sets the stage for the multi-
country model to come. This is the model in Dornbusch, Fischer, and Samuelson (1977).
The model is static. The economy consists of two countries, country 1 and country 2.1 Each
country is endowed with labor, `¯1 = `¯2 . Technologies exist to produce a continuum of goods,
indexed by k ∈ [0, 1], but these technologies differ across countries. There are no costs to trading
in this version of the economy; we will add them below. The technology to produce good k in
country i is
1
yi (k) = `i (k), (1)
ai (k)
where ai (k) is the unit labor requirement for producing good k in country i. yj (k) is the output of
good k produced in country i and `i (k) is the amount of labor used in production. Assume that
the unit labor requirement functions are

a1 (k) = exp (α (1 − k)) (2)

a2 (k) = exp (αk) . (3)

The functional form for ai can be quite general. What is important is that we can rank each good
according to the relative productivity of producing it in each country. For a two-country economy,
this is simple — we will do it graphically below — but in a multicountry world, this is nontrivial.
The representative consumer in each country chooses consumption of each good to maximize
Z 1
log (ci (k)) dk (4)
0

subject to the budget constraint,

Z 1
pi (k)ci (k) dk = wi `¯i (5)
0

and the constraint that consumption is nonnegative. Firms are perfectly competitive. The repre-
1
This example is based on one from Tim Kehoe, http://www.econ.umn.edu/˜tkehoe/teaching.html

Subject to change. This version December 21, 2016. |1

Ricardian Models Kim J. Ruhl

sentative firm in country i producing good k solves

max p(k)yi (k) − wi `i (k) (6)

`i (z)
1
s.t. `i (k) ≥ yi (k).
ai (k)

Market clearing has total consumption equal to total output, labor supply equal to labor demand
in each country, and trade balance

y1 (k) + y2 (k) = c1 (k) + c2 (k), k ∈ [0, 1] (7)

Z 1
`i (k) dk = `¯i , i = 1, 2 (8)
Z 0 Z
p1 (k)c1 (k) dk = p2 (k)c2 (k) dk, (9)
K21 K12

where Kij is the set of goods that are produced in country i and consumed in country j, i.e., the
goods exported from i to j.
Let pi (k) be the price of good z in country i. An equilibrium of this model is prices {w1 , w2 , p1 (k), p2 (k)}
and quantities {`1 (k), `2 (k), c1 (k), c2 (k), y1 (k), y2 (k)} such that households and firms solve their
maximization problems and markets clear.
The symmetry of the problem implies that an equilibrium exists in which w1 = w2 = 1. What is
the pattern of trade? The household in country 1 will purchase good z from whichever producer
provides it at the lowest cost, so the price of the good in country i is

pi (k) = min {p1 (k), p2 (k)} . (10)

From the producer’s problem in (6), the price of the good produced in country i is pi (k) = wi ai (k).
So country 1 will produce the good for consumption in both countries 1 and 2 whenever

a1 (k)w1 < a2 (k)w2 (11)

a1 (k) w2
< , (12)
a2 (k) w1

and country 2 will produce k for consumption in both countries when the opposite inequality is
true. Let’s plot the two parts of (12) in figure 1.
To find k̂, the cutoff good, solve
    
w1 exp α 1 − k̂ = w2 exp αk̂ , (13)

which, in this symmetric world, yields k̂ = 0.5. The equilibrium prices are
(
w2 exp(αk) k ∈ [0, k̂]
p1 (k) = p2 (k) = (14)
w1 exp(α(1 − k)) k ∈ (k̂, 1]

Subject to change. This version December 21, 2016. |2

Ricardian Models Kim J. Ruhl

Figure 1: The pattern of trade.

w2
w1

a1 (k)
a2 (k)

Country 2 exports Country 1 exports k

these goods k̂ these goods

consumption is
wi `¯i
ci (k) = i = 1, 2 (15)
pi (k)
and production and employment are
¯ ¯
y1 (k) = 0 `1 (k) = 0, y2 (k) = w1 `p(k)
1 +w2 `2
`2 (k) = 2`¯2 k ∈ [0, k̂]
w1 `¯1 +w2 `¯2 (16)
y1 (k) = p(k) `1 (k) = 2`¯1 , y2 (k) = 0 `2 (k) = 0 k ∈ (k̂, 1].

The symmetric case is very easy to solve. The nonsymmetric case is pretty simple, too. In the
nonsymmetric model, set w1 = 1 and solve (13) and (9) for w2 and k̂.
Two countries and a continuum of goods with tariffs
Assume each country imposes a tariff of τij on imports from the other country. The tariff revenue
is lump-sum rebated to the household. The rest of the model is the same as the previous model.
The new budget constraint is
Z Z
pii (k)ci (k) dk + pjj (k)(1 + τij )ci (k) dk = wi `¯i + Ti (17)
Kii Kji

and the tariff revenue conditions are

Z
Ti = τij pjj (k)ci (k) dk. (18)
Kji

The first integral in (17) is the expenditure on goods produced domestically and the second in-
tegral is import expenditure. Additionally, the balanced trade condition should have trade flows
valued at pre-tariff prices. (In the symmetric world this doesn’t matter, but in a nonsymmetric

Subject to change. This version December 21, 2016. |3

Ricardian Models Kim J. Ruhl

world it does.)
What is the pattern of trade in this model? For country 2 to import a good from country 1,

w1 a1 (k)(1 + τ12 ) < w2 a2 (k) (19)

must be true, and for country 1 to import the good from country 2, it must be that

w1 a1 (k) > w2 a2 (k)(1 + τ21 ). (20)

As we did in the previous model, we can plot these conditions.

Figure 2: Tariffs generate a range of nontraded goods.

w2 (1+τ21 )
w1

w2
w1 (1+τ12 )

a1 (k)
a2 (k)

Country 2 exports Nontraded Country 1 exports k

these goods kb2 goods kb1 these goods

The tariffs generate a range of goods, (kb2 , kb1 ), such that

w2 a1 (k) w2 (1 + τ21 )
< < k ∈ [kˆ2 , kˆ1 ]. (21)
w1 (1 + τ12 ) a2 (k) w1

These are goods for which the comparative advantage of production is not strong enough to over-
come trade costs. The goods are not exported by either country, and we have incomplete special-
ization in this economy. [The slope of the relative productivity curve summarizes the strength of
comparative advantage in the world.]
In the symmetric case, solving for equilibrium requires solving (19) and (20) (with equality im-
posed) for kb1 and kb2 . In the nonsymmetric world, add w2 and the balanced trade condition.

2 The multicountry world

The Dornbusch, Fischer, and Samuelson (1977) model’s elegance follows from the ability to sort
the goods according to the comparative advantage of one country relative to another. With only
two countries, we can always “rename” the goods to satisfy this ordering. How do we generalize
this ordering to a world with more than two countries? In a deterministic environment, this is a

Subject to change. This version December 21, 2016. |4

Ricardian Models Kim J. Ruhl

difficult task. Wilson (1980) is a notable study.

Eaton and Kortum (2002) shows that the multicountry Ricardian model can be quite tractable
when productivity is random. For any good k, the set of prices offered by the countries of the
world becomes a random variable, and with a judicious distributional assumption for productiv-
ity, the model is extremely tractable.
Two kinds of results here. 1) Making productivity stochastic makes the multicountry model work.
2) Guessing Frechet makes the thing closed form. We can abandon 2) and 1) still works great, we
just need a computer to solve the model.
Technologies
As in Dornbusch, Fischer, and Samuelson (1977), there are an continuum of goods, indexed k ∈
[0, 1]. The technology to produce a good is given by

yi (k) = zi (k)`i (k), (22)

where z — the marginal productivity — is now a random variable. Producers are competitive and
all producers of good k in country i receive the same productivity draw, zi (k). Goods are subject
to iceberg trade costs, τij , with the normalization that τii = 0.
The productivity draws, z, are distributed according to a Type II extreme value distribution, or a
Frechet distribution,
−θ
Fi (z) = prob (zi (k) ≤ z) = e−Ti z . (23)

The parameter Ti > 0 controls the mean of the distribution and the parameter θ > 1 controls the
dispersion in the distribution. Notice that T is country specific: We will allow countries to have
different average levels of productivity. In these notes θ is identical across countries because it
allows for easy aggregation, but this is not necessary if you are willing to use a computer to solve
the model.2
The productivity draws are i.i.d. across goods and countries. Again, we can relax this assumption,
but, again, we would then need a computer to solve the model.
The functional form assumption is not completely ad hoc. The Frechet distribution is one of the
limiting distributions that results from sequentially drawing numbers and keeping the best draws.
This can be thought of as innovation: Try some things (take a draw) and keep it if it is better than
what you already have.
The parameter θ is an important part of this model. By controlling the amount of dispersion in
productivity levels, it controls the potential gains from trade: In Ricardian models, countries trade
to take advantage of differences in the cost of producing a particular good across countries. If
countries are very similar (low dispersion) there won’t be many opportunities to trade.
2
This is a one sector model. See Caliendo and Parro (2012) for a multisector model with sector-specific θ.

Subject to change. This version December 21, 2016. |5

Ricardian Models Kim J. Ruhl

Preferences
The representative agent in country n in endowed with `¯n units of labor, and has preferences over
consumption,
Z 1 σ
 σ−1
σ−1
Un = cn (k) σ dk . (24)
0

Taking as given the entire set of prices offered by each country for each good, agents in country n
choose cn (k) to maximize utility subject to the budget constraint,
Z 1
pn (k)cn (k) ≤ wn `¯n . (25)
0

Market clearing
Market clearing is the same as in the two country world: each good market (k) clears, the labor
market in each country clears, and trade balances for each country. Note that trade does not
balance bilaterally, each balances in the aggregate for each country.
Intuition
At this point, it is worth thinking through how this model works before embarking on the calcu-
lus that follows. Suppose we discretize the model (this is what we will be doing on the computer,
anyways). Let there be 1,000,000 goods in the economy. Each country draws 1,000,000 productiv-
ities from their distribution, one for each good. Each country’s distribution is characterized by θ
and Ti . With the productivity in each country and each good pinned down, we can compute the
price any country would charge another for any good k, and from there, we can compute which
country will sell to which country. (It’s the one with the lowest price.) From there, we just need to
find the relative wages that clear markets.
When we use the Frechet distribution for the productivity distribution, we can compute closed-
form solutions for the trade flow between any two countries. This is the value added of this
distributional assumption. With the closed-form solutions, we can gain intuition about how the
model works.
Analysis
From here onward, we are going to make a change in notation that is common in these kinds of
models. Instead of keeping track of goods by their names (k) we are going to keep track of goods
according to the value of the productivity draw (z = z(k)) the good receives. We do this because
it is the productivity value that matters for the firm, not the actual name of the good.

Significant departure from the previous notation. To be consistent with Eaton and
Kortum (2002), I will use xni to be the value of x from i to n, the opposite ordering
of the indices in the other notes.

Subject to change. This version December 21, 2016. |6

Ricardian Models Kim J. Ruhl

The problem for the firm producing a good with productivity z in i for sale in n is

max pni (z)y(z) − wi `i (z) (26)

`i (z)

s.t. y(z)(1 + τni ) = z`i (z) (27)

which yields the pricing condition,

wi
pni (z) = (1 + τni ) . (28)
z

As in the two-country model, the consumer will choose to purchase goods from the lowest priced
seller. The multicountry version of (10) is

pn (z) = min {pn1 , pn2 , . . . , pnN } (29)

 
w1 (1 + τn1 ) w2 (1 + τn2 ) wN (1 + τnN )
pn (z) = min , ,..., (30)
z z z
 
z z z
pn (z) = max , ,..., (31)
w1 (1 + τn1 ) w2 (1 + τn2 ) wN (1 + τnN )

What do we want to know? We want to know how much country n buys from country i.

1. What prices does i offer to n, Gni (p)?

2. [What are the best prices offered to n, Gn (p)?]

3. Given all of the Gnj , what is the probability that i is the lowest cost supplier, πni ?

4. By the l.l.n., πni is also the share of the commodity space that i sells to n. At this point we
know that n buys a share of the goods from i, but we don’t know the value of this trade yet.

5. Show that the price distribution of the goods i sells to n is identical to the price distribution
of goods sold to n by every other country, Gn . This result implies that, πni is also the share
of total spending by n on goods from i. This is what we wanted to know!

Okay, let’s get started. The distribution of prices that country i offers country n is
 
wi (1 + τni )
Gni (p) = prob <p (32)
z
 
wi (1 + τni )
Gni (p) = prob <z (33)
p
 
Gni (p) = 1 − exp −Ti (wi (1 + τni ))−θ pθ . (34)

Since the productivity draws are i.i.d. across goods and countries, the distribution of prices over
the goods that the country n agent will actually buy is

Gn (p) = prob (pn < p) = 1 − ΠN

n=1 (1 − Gni (p)) (35)

Huh? Let’s break this down into pieces.

Subject to change. This version December 21, 2016. |7

Ricardian Models Kim J. Ruhl

• Gni (p) is the probability that i sells the good for less than p, so
• 1 − Gni (p) is the probability that i sells to n at greater than p.
• The i.i.d. draws mean that ΠN
n=1 (1 − Gni (p)) is the probability that every country sells to n
at price greater than p, so
• 1 − ΠNn=1 (1 − Gni (p)) is the probability that at least one country sells the good to n for less
than p.

Now use the (34) to yield

  
−θ θ
Gn (p) = 1 − ΠNn=1 1 − 1 + exp −Ti (wi (1 + τni )) p (36)
N
!
−θ θ
X
Gn (p) = 1 − exp −Ti (wi (1 + τni )) p (37)
i=1
 
Gn (p) = 1 − exp −Φn pθ , (38)

which is also a Frechet distribution, where

N
Ti (wi (1 + τni ))−θ ≥ Tn wn−θ
X
Φn = (39)
i=1

The Φ term summarizes the forces that shape trading opportunities in the economy. Country n
is able to trade with countries that vary in their their levels of productivity (Ti ), their factor costs
(wi ), and their trading costs (τni ). This term should remind you of the multilateral resistance terms
that we encountered in Anderson and van Wincoop (2003).

• If trade costs are infinite with everyone but your own country Φn = Tn wn−θ and your price
distribution is just Gnn (p), which is determined by only country n’s technology. This is
autarky. Trading with at least one partner gives us Φn > Tn wn−θ so trading gives you access
to a better distribution of prices. It’s as if country n has better technology once it has trading
partners.
• If trade costs are zero, then Φ is the same everywhere, and every country will buy each good
from the same producing country. This is complete specialization, just as in the two country
model without trade costs.
• The greater is θ, the more sensitive is the price distribution to trade costs. Notice that, in this
way, θ plays a role similar to the elasticity of substitution in the Armington model.

Let’s prove some important properties of the price distribution.

The probability that a country is the lowest cost supplier
Let πni be the probability that country i is the lowest cost supplier of good k to country n,
Z ∞
πni = prob (pni (k) ≤ pns (k) ; s 6= i) = ΠN
s6=i [1 − Gns (p)] dGni (p) (40)
0

where

Subject to change. This version December 21, 2016. |8

Ricardian Models Kim J. Ruhl

• dGni (p) is the probability that i offers price p

• 1 − Gns (p) is the probability that s offers a price greater than p

• ΠN
s6=i [1 − Gns (p)] is the probability that every other country offers a price greater than p.

• Integrate this over all possible prices, p ∈ (0, ∞)

With
 
• dGni (p) = θTi (wi (1 + τni ))−θ pθ−1 exp −Ti (wi (1 + τni ))−θ pθ
P 
−θ θ
• ΠN
s6=i [1 − G ns (p)] = exp s6=i −Ts (w s (1 + τns )) p

This makes the integral, (then multiply by Φn /Φn )

Z ∞ n o
πni = Ti (wi (1 + τni ))−θ θ exp Φn pθ pθ−1 dp (41)
0
Z ∞
1 n o
πni = Ti (wi (1 + τni ))−θ θΦn exp Φn pθ pθ−1 dp (42)
Φn 0

The term inside the integral is the pdf of a Frechet distribution. It’s integral is equal to 1, so

Ti (wi (1 + τni ))−θ

πni = . (43)
Φn

By the law of large numbers, this probability is also the fraction of goods that country i actually
sells to n. Here we see more multilateral resistance stuff: The i to n bilateral trade depends on i’s
average productivity, factor costs, and bilateral trade costs relative to these factors in every other
country.
The price distribution of imports
At this point, we know what share of goods country i sells to country n, πni . We do not know, yet,
the value (p × q) of that trade. To do so, we need to know: What does the price distribution look
like, conditional on the goods being bought are from i?
Z p
1
prob (pn < p | pn = pni ) = Πs6=i [1 − Gns (q)] dGni (q) (44)
πni 0
Ti (wi (1 + τni ))−θ p
Z
Φn n o
= θΦn exp Φn q θ q θ−1 dq (45)
Ti (wi (1 + τni ))−θ Φn 0
 
= 1 − exp −Φn pθ (46)

The last expression is Gn (p): The price distribution of the goods that country n actually buys from
i is identical to the price distribution of the goods it buys from every other country. This implies
that the average price of the goods shipped from i to n is the same as the average price of the
goods shipped from j to n. This means that πni is not only the share of the goods that i ships to n,
but that it is also the share of country n’s expenditure that is spent on goods from i.

Subject to change. This version December 21, 2016. |9

Ricardian Models Kim J. Ruhl

Note that since average prices are equal across importers, better importers — those with higher
technology levels, lower factor prices or lower bilateral trade barriers — sell more goods to coun-
try n, and not more expensive goods, which we see from (43).
Gravity
Since the price distributions of the goods actually purchased by n are identical, the average price
of the goods purchased are identical, so the share of spending by n on goods from i is just equal
to the share of goods purchased from i, πni . Let Xni be spending (p × q) by n on goods from i

Xni Ti (wi (1 + τni ))−θ

= πni = (47)
Xn Φn

Total sales by country i is

N N N
X X X (1 + τmi )−θ
Qi = Xmi = πmi Xm = Ti wi−θ Xm (48)
Φm
m=1 m=1 m=1

solve for Ti wi−θ and substitute it into the previous equation

(1 + τni )−θ 1
Xni = PN   Xn Qi (49)
Φn −θ −1
m=1 (1 + τmi ) Φm Xm

−1/θ
use Pn = γΦn to substitute out the Φ terms3
 −θ
1+τni
Pn
Xni =  −θ Xn Qi (50)
PN 1+τmi
m=1 Xm Pm

and we have a gravity equation. Some interesting things:

• The country incomes enter with unit elasticities, as in Anderson and van Wincoop (2003)
• The price level in n deters trade from i. The lower is the price level in n, the stiffer is compe-
tition, making it less likely i is a lowest cost supplier.
• θ plays the role that σ did in the Armington model. A larger θ implies more heterogeneity.
Remember that each country that sells to n does so with the same distribution of prices, so
there is little room for the elasticity of substitution to come into play. A country pair with
a high trade volume is a country pair that trades a lot of goods on the extensive margin. θ
determines how the extensive margin works.

A first look at the gains from trade

Welfare in a country is summarized by the real wage,
wi wi
= −1/θ
. (51)
Pi γΦi
3
This is result (c) on page 1748. The γ term is a constant that involves θ and σ. This is the only place that σ shows
up: We can ignore σ in the Frechet case.

Subject to change. This version December 21, 2016. | 10

Ricardian Models Kim J. Ruhl

The definition of π give us

Ti (wi )−θ
πii = (52)
Φi
−1/θ −1/θ 1/θ
Φi = Ti wi πii (53)

substitute to get
wi −1/θ 1/θ
= γ −1 πii Ti . (54)
Pi
The gains from trade depend on parameters and how much a country purchases from itself. In
−1/θ
autarky, πii = 1, so the gain from moving from autarky to trade is (wi /pi )/(wiA /pA
i ) = πii ,
where variables with superscript A are those in the autarky equilibrium. Example: if θ = 4 and
πii = 0.9 then the gain from moving from autarky to a world where country i imports one-tenth
of its total expenditure is 1.027, or 2.7 percent.
Parameterization
In the Eaton and Kortum (2002), production involves labor and intermediate goods, and there are
nontraded goods. These features add parameters β, the share of labor in production, and α, the
share of traded goods in total output. We are abstracting from these here, but they are important
for welfare analysis. Liberalizing trade, for example, only has a direct impact on the traded good
sector, which is relatively small compared to the nontraded sector.
We need: θ, Ti , dni .
Data: bilateral trade flows, Xni and wages, wi

Estimating θ The procedure used in the paper is problematic. See Simonovska and Waugh (2014)
for a discussion and estimation. They find that θ = 4. That’s become the “standard” value,
but keep in mind that parameters are specific to a model.

Estimating geography and technology Using (47) for both n to i trade and n to n trade, we have
 −θ
Xni πni Ti wi
= = (1 + τni ) . (55)
Xnn πnn Tn wn

Take logs  
Xni Ti wi
log = −θ log(1 + τni ) + log − θ log . (56)
Xnn Tn wn
Let Si = log Ti − θ log wi , which is a “competitiveness” measure: the technology parameter
adjusted for the wage level. Now we have

Xni
log = −θ log(1 + τni ) + Si − Sn , (57)
Xnn

which we can estimate, with country fixed effects. How do we deal with geography? Pa-
rameterize the trade costs as

log(1 + τni ) = dk + b + ` + eh + mn + δni , (58)

Subject to change. This version December 21, 2016. | 11

Ricardian Models Kim J. Ruhl

where dk are dummy variables associated with distance “bins”, b is a border dummy, ` is a
language dummy, eh is an RTA dummy, mn is a destination effect. Now we estimate

Xni
log = −θ{dk + b + ` + eh + mn + δni } + Si − Sn , (59)
Xnn

Note that we are estimating θ times the structural parameters. We will use the value of θ to
back out the underlying parameters.
Estimates are in table III: Distance matters; borders not much; Japan and USA are most
competitive, Belgium and Greece the least.
Table VI has the backed out parameters. This table allows us to get a sense of technology
versus factor costs. For example, Japan’s S is larger than the USA’s, but that reflects low
wages in Japan: TU S > TJP N .

Counterfactuals
With the estimated parameters, we can now use the model to perform counterfactual experiments.
Eaton and Kortum (2002) consider two kinds of labor markets: an immobile one, where labor can-
not move across the traded and nontraded sectors, and a mobile one, where labor can be reallo-
cated. A change in welfare in the model boils down to the change in the real wage (country size is
constant), wi /pi .

1. Autarky. Let τni = ∞ for all n 6= i. Table IX. Modest decreases in welfare everywhere.
Belgium: not very competitive, but close to lots of competitive countries. They lose a lot.

2. Weightless trade. Let τni = 0 for all n, i. Table X. Much larger welfare gains. The real world is
much closer to autarky than to free trade. Notice how small countries that are far away (New
Zealand, Norway, Portugal) increase their traded sector employment. These were countries
whose technology was pretty good, but faced high geographic barriers.

3. A more realistic exercise might by to decrease tariffs to the point that world trade has dou-
bled: a 69 percent decrease. Table X. Gains are similar to those in the autarky case.

4. Increase technology in a country and look for welfare implications. Table XI. Spillovers are
largest in countries that are close: Canada gains the most from increase in USA technology.
Austria, Belgium, and Denmark from an increase in German technology.

Computation
Let N be the number of countries, with Ti , θ, σ, and τni given as parameters. Discretize the interval
of goods: Let K be a large number, say 100,000.

1. Initialize the program. For each i = 1, . . . , N draw k = 1, . . . , K productivities from country

i’s distribution.

2. Guess a vector of N − 1 wages. (Normalize w1 = 1.)

3. For each importing country n

Subject to change. This version December 21, 2016. | 12

Ricardian Models Kim J. Ruhl

(a) For each good k compute the price from each potential exporter, pni = wi /zi (k)τni .
(b) Find the minimum delivered price for each good, and note the supplier. The result will
be a K-vector of prices and a K-vector of the associated supplying countries.
(c) Construct a K-vector of spending on each good, pn (k)cn (k).

4. Aggregate the goods-level trade flows to country aggregates. You should have an N × N
matrix of trade flows.

5. Check: Does trade balance? If it does, you have found an equilibrium. If not, return to step
2. (You only need to check N − 1 balanced trade conditions. You get one for free from Walras
Law.) Notice that this is a system of N − 1 equations in N − 1 unknowns, the wages.

If we let the T and the τ all be equal, then we have a perfectly symmetric model. Let the wage be
1 in every country, and trade should balance without having to adjust the wages. If it does not,
you may have a bug in the program or you may need to increase J.

Subject to change. This version December 21, 2016. | 13

Ricardian Models Kim J. Ruhl

——————
Anderson, James E. and Eric van Wincoop (2003). “Gravity with Gravitas: A Solution to the Border
Puzzle.” American Economic Review 93 (1), pp. 170–192.
Caliendo, Lorenzo and Fernando Parro (2012). “Estimates of the Trade and Welfare Effects of
NAFTA.” NBER Working Papers 18508.
Dornbusch, Rudiger, Stanley Fischer, and Paul A Samuelson (1977). “Comparative Advantage,
Trade, and Payments in a Ricardian Model with a Continuum of Goods.” American Economic
Review 67 (5), pp. 823–39.
Eaton, Jonathan and Samuel S. Kortum (2002). “Technology, Geography, and Trade.” Econometrica
70 (5), pp. 1741–1779.
Ricardo, David (1817). On the Principles of Political Economy and Taxation. London: John Murray.
Simonovska, Ina and Michael E. Waugh (2014). “The Elasticity of Trade: Estimates and Evidence.”
Journal of International Economics 92 (1), pp. 34–50.
Wilson, Charles A. (1980). “On the General Structure of Ricardian Models with a Continuum of
Goods.” Econometrica 48 (7), pp. 1675–1702.

Subject to change. This version December 21, 2016. | 14